agent-based coordination scheme for pv ......shines program review 2017 energy.gov/sunshot...

energy.gov/sunshotenergy.gov/sunshotSHINES Program Review 2017

energy.gov/sunshotenergy.gov/sunshotenergy.gov/sunshot

SHINES Program Review 2017

AGENT-BASED COORDINATION SCHEME FORPV INTEGRATION (ABC4PV)Awarded to CMU, NRECA & Aquion Energy

Presented by Panayiotis Moutis, PhD (CMU)

energy.gov/sunshotenergy.gov/sunshotSHINES Program Review 2017SHINES Program Review 2017 2

• Motivation

• Technical Objectives

• Team

• Approach

• Anticipated Outcomes

• Project Plan, Outcomes & Milestones Achieved in 2016

• Discussion

Presentation Outline


• Increase penetration of photovoltaics, storage &

demand response, jointly operating based on . . .

• Distributed (scalable) optimal (cost effective) control

Motivation

Gap between theoretical approaches and actual application

Need for a testbed for distributed algorithms, a testbed for the application of PV integration and

a testbed of such algorithms in a practical environment


• Define proper set of active assets (PV, storage, loads) = “Unit”

• Extend & Test existing distributed optimal control algorithm

• Extend to multiple assets per control point (PV, storage, switches)

• Test for stability, robustness & malicious incidents

• Verify the real-world operation of 10 units on test-bed

• Install units on end-customer households of a co-op feeder

• Define testing protocols and cyber tools

• Determine communication architecture (due to algorithm)

• Assess algorithm optimality (due to extensions & framework)

• Study impact of storage system (notably, lifetime) on optimality

Technical Objectives


Team

• CMU (S. Kar, G. Hug, J. Moura, P. Moutis, J. Whitacre)

• Operating control framework

• Efficacy assessment (simulations & software implementation)

• NRECA (C. Miller, A. Cotter, D. Danley, D. Pinney)

• Testbed development

• Standardization of unit equipment

• Aquion Energy (T. Madden, J. Whitacre)

• Battery storage system physical integration and optimization

energy.gov/sunshotenergy.gov/sunshotSHINES Program Review 2017energy.gov/sunshotenergy.gov/sunshot6

ABC4PV

…

Roadmap Approach:- Determine “unit”- Inter-unit collaboration


Approach: Unit-based integration and abstraction

• The ABC4PV unit

• Max integration & combinational value

• Plug-n-playcapabilities

• Assuming on high penetration (i.e. a large number of “units” has been deployed)


Approach: Inter-unit collaboration

• Agent-based distributed optimal control algorithm:

• Units shareinformation

• System operator optional

• Local calculations


Approach: Distributed optimization framework

• Form of updates

– Mathematical formulation based on Consensus + Innovation

approach:

consensus term: Lagrangian multiplier representing the marginal cost

innovation term: -»- -»- for global constraint(s)

𝜆𝑗(𝑖+1)

= 𝜆𝑗(𝑖)− 𝛽𝑖

𝑙∈𝜔𝑗

𝜆𝑗(𝑖)− 𝜆𝑙

(𝑖)

𝑐𝑜𝑛𝑠𝑒𝑛𝑠𝑢𝑠

− 𝛼𝑖 መ𝑑𝑗(𝑖)

𝑖𝑛𝑛𝑜𝑣𝑎𝑡𝑖𝑜𝑛

𝑗=1

𝐽

መ𝑑𝑗(𝜆𝑗) = 0

Low computational effort required

Considered Application:

- Consensus converges to cost optimality

- Innovation effect fades with convergence


• Key Project Products (specific “artifacts”)

• Framework for a maximum value, plug’n’play unit design

• Scalable and robust distributed algorithm for optimal control

• Test-bed verified guarantees

• Major Outcomes (conceptual achievements)

• Pathway towards efficient integration of up to 100% solar (or distributed generation) penetration on optimal control

• Functioning, adjustable & expandable test-bed for distributed algorithms

Anticipated Outcomes


• Distributed Control Methodology• Design and Development of Distributed Resource Coordination and

Control Methodology (Years 1, 2)

• Software Implementation of Algorithms (Years 1, 2)

• Cybersecurity Assurance (Years 1, 2, 3)

• Testbed Development• Installation of Physical Components and Test-bed Setup (Years 1, 2)

• Performance Testing in Test-bed (Years 2, 3)

• Functional and Economic Assessment and Optimization• Value of Solar (Years 2, 3)• Value of Storage (Years 2, 3)• Energy Storage Sub-System and Full System Assessment and Control

Optimization (Years 2, 3)

Project Plan


Distributed Control Methodology Development


Subtasks

1. Modeling of Cost/Objective

• Objective function of the optimization problem

2. Modeling of Constraints (Operational and Uncertainty)

• Component operation constraints, e.g. min/max inverter power

• Network constraints, e.g. line loading limits

3. Formulation of Iterative Distributed Control

• Employ optimization algorithm to the problem

4. Performance Analysis and Real-time Guarantees

• Benchmarking of convergence and time complexity

5. Simulation of Test-bed

Task 1: Design and Development of Distributed Resource Coordination and Control Methodology


Objective Function

]24,1[

]24,1[}{

,,

}{

,,

}{

,,

}{

,, tt

T

tbatteriesi

tDChi

PVi

tNMi

batteriesi i

tRChi

loadsi

tLi IFPPn

PP

]24,1[

]24,1[}{

,,

]24,1[ }{

,, tt

T

tloadsi

tILi

t batteriesi

tRChi ILCPDgd

]24,1[,

]24,1[}{

,,

ttPV

T

tPVi

tCntri CntrP

minimize

• Expresses total cost of coop feeder load demand energy, subtracting PV, storage & demand response contributions


Problem Constraints

Operating, system, policy, etc. . .

. . .

. . .

. . .

Non-convex power flow equality constraint

iii vvv ijijij SSS

***

ijjiiijij yvvvqip and j

iji pP j

iji qQand


Elaboration on Non-Convex Power Flow Constraint

• Relaxations lead to computationally tedious algorithms

• Existing linear approximations (DC & Decoupled) fail to consider the resistive nature of the distribution systems

• Two novel “resistive-aware” linear approximations developed


Novel linear OPF approximation indicative results

- 10-bus radial distribution feeder on ACSR-95 lines (9 loads & DG units)

Decoupled OPF Novel linearized OPF SDP-relaxed OPFV in p.u. Angle (rad)

1.0000 0

0.9959 -0.0056

0.9792 -0.0281

0.9762 -0.0322

0.9750 -0.0338

0.9731 -0.0365

0.9723 -0.0377

0.9708 -0.0398

0.9703 -0.0406

0.9700 -0.0410

V in p.u. Angle (rad)

1.0000 0

0.9949 -0.0034

0.9744 -0.0171

0.9707 -0.0196

0.9693 -0.0205

0.9669 -0.0222

0.9658 -0.0230

0.9640 -0.0242

0.9633 -0.0247

0.9629 -0.0249


1.0000 -0.0000

0.9950 -0.0024

0.9743 -0.0125

0.9705 -0.0144

0.9691 -0.0151

0.9667 -0.0164

0.9655 -0.0170

0.9637 -0.0180

0.9631 -0.0184

0.9626 -0.0185


Novel linearized OPF approximation in C+I f/w results

Novel central Novel on C+I

Compared to the Decoupled OPF on C+I, novel OPF 30-60% faster


1.0000 0

0.9949 -0.0034

0.9744 -0.0171

0.9707 -0.0196

0.9693 -0.0205

0.9669 -0.0222

0.9658 -0.0230

0.9640 -0.0242

0.9633 -0.0247

0.9629 -0.0249


1 0

0.99711 -0.0033843

0.98521 -0.017112

0.98302 -0.019603

0.98226 -0.020528

0.98087 -0.022189

0.98024 -0.022955

0.97925 -0.024232

0.97886 -0.024679

0.97860 -0.024934


Brief note on the novel Battery operating cost model

Unlike simplified linear cost models of Battery Storage systems, a novel cost model has been developed

• Effect of CapEx depreciation accounted for

• Battery lifetime (throughput) expressed as a function of “usability”

More precise consideration of battery operating cost for similar control problems…


Subtasks

1. Network Design

• Assessment of the selected coop feeder as the test bed

2. Implementation of Code on Embedded Site Controller and Communication

• Validate performance of operation framework in simulation

• Validate performance of operation framework in situ

3. Write Host Program to Collect and Format Data

Task 2: Software Implementation of Algorithms


Indicative power flow output on CoServ feeder


Effect of distribution feeder design on power quality

• Ikaria island (Greece)

R-22 feeder on peak load

• Rhodes island (Greece)

R-22 & R-26 feeders on

installed (not peak) load

• Ikaria within power quality standard, Rhodes not (at various levels >50% installed)

• Distribution Networks usually oversized…


Subtasks

1. Security Design

• Analysis of cyber-security concerns and suggested measures

2. Secure Software Testing

• Testing cyber-security performance of developed framework

3. System and In-Situ Testing

• Cyber-security testing on the test-bed implementation

Task 3: Formulation of Iterative Distributed Control


Testbed Development


Subtasks

1. Identification of Suitable Feeder

• Promote project to candidate coop feeders

2. Design/Sizing of Devices

• Determining exactly the components of each unit

3. Initial Deployment and Validation

• Assessment of unit components (inter-)operability

4. Full Deployment and Validation

• Unit fine-tuning based on cost assessment and final installation

Task 4: Installation of Physical Components and Test-bed Setup


Unit Design – Bill of Materials

PV Subsystem

Item Description Manufacturer Part #

1.1 PV Module Suniva 260 Wp

1.2 Racking Rooftech RT-[E]

1.3 Optimizer Tigo Tigo TS4L

1.4Optimizer Comms / Rapid Shutdown Tigo Gateway

1.5 Arc Fault Detection DC SunvoltADU - Arc Fault Detection Module

Battery Subsystem


2.1 Battery Aquion Aspen 48M

2.2 Battery BMS Aquion

Power Electronics Subsystem


3.1 Inverter Schneider Conext XW+ 5548-NA

3.2Power Distribution Panel Schneider Conext XW+ PDP

3.3 Charge Controller Schneider XW-MPPT80-600

3.4 Control Panel Schneider Conext 865-1050-01

3.5 Comms Panel Schneider Conext 865-1058 ComBox

3.6 Battery Monitor Schneider Conext 865-1080-01

3.7 100A DC Breaker Schneider 865-1070

Controls Subsystem


4.1 Whole House Monitor RCS Model EM52-ZW

4.2 Thermostat RCS Model TZ45

4.3 Load Control Relay Evolve Evolve LFM-20

4.4 Load Power Contactor ElkHeavy Duty Relay Contactor

4.5 Zwave Gateway Vera VeraEdge Z-Wave Controller

4.6 System Controller w/ enclosure tbd


Unit Design – One line Diagram


Unit equipment procurement & testing set-up


Formally quantified results at the end of 2016

• Average LCOE below $0.14/kWh (seasonal can spike)

• Additional analysis to follow

• Power flow simulations on coop feeders (and others) determine dimensioning as main power quality factor

• Cybersecurity assurances procured and testing protocols determined

• Test bed (coop feeder) identified (could be revisited)

• Unit components, topology and design finalized

• Recharging/discharging tests successful

• Control & optimal algorithm operation tested


Dissemination actions

• Conference paper (S. Weerakkody, B. Sinopoli, S. Kar, and A.

Datta, “…,” IEEE Conference on Decision and Control, 2016)

• Journal submission (C. Wu, G. Hug, and S. Kar, “…,” Submitted to

IEEE Transactions on Smart Grid, Oct. 2016)

• P. Moutis’ invited talks at Princeton Uni., CalTech, UCLA, UIUC

• Power engineering letter (P. Moutis, G. Hug, and S. Kar, “…,”

Submitted to IEEE Transactions on Power Systems, Nov. 2016)

• Conference publication (C. Wu, G. Hug, and S. Kar, “…,”

American Control Conference (To appear), May 24–26, 2017, Seattle,

WA)


Questions?

Comments?

agent-based coordination scheme for pv ......shines program review 2017 energy.gov/sunshot...

Documents